[FEATURE] Permission popups: show a plain-language description of what the command does, not only the raw command
Motivation
When Claude Code wants permission to do a dangerous or human-gated action, it shows a popup with the command it is going to run.
That command is almost always unreadable, or reading it would take minutes.
If the popup showed what the command does, instead of the raw command, it would greatly increase readability, security and user trust.
Instead of sed xyz$$^ --ignore --param111 "asidh --alpha", we would see "Try to get a new token from Amazon", which we can approve more wholeheartedly.
It could also bring back an "I love to live dangerously" mode (just yes-yes-yes in the background) to save time.
Suggestion
Show a plain description of what the command actually does in place of the raw command. Two ways to get it:
- Have the response already return the description: the agent writes it together with the tool call.
- Have the UI send the command string to a cheap or near free model, and show what it returns.
The extra latency is not noticeable compared to the time of the agent's work. The time saved by actually being able to read the popup is larger than the ~1 second the API call takes in the worst case.
Optional 1: a collapsible section to open and see the original command.
Optional 2: a setting for which one is shown at the top: the raw command, or the plain description.